2 * linux/arch/i386/kernel/setup.c
4 * Copyright (C) 1995 Linus Torvalds
6 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
8 * Memory region support
9 * David Parsons <orc@pell.chi.il.us>, July-August 1999
11 * Added E820 sanitization routine (removes overlapping memory regions);
12 * Brian Moyle <bmoyle@mvista.com>, February 2001
14 * Moved CPU detection code to cpu/${cpu}.c
15 * Patrick Mochel <mochel@osdl.org>, March 2002
17 * Provisions for empty E820 memory regions (reported by certain BIOSes).
18 * Alex Achenbach <xela@slit.de>, December 2002.
23 * This file handles the architecture-dependent parts of initialization
26 #include <linux/sched.h>
28 #include <linux/mmzone.h>
29 #include <linux/screen_info.h>
30 #include <linux/ioport.h>
31 #include <linux/acpi.h>
32 #include <linux/apm_bios.h>
33 #include <linux/initrd.h>
34 #include <linux/bootmem.h>
35 #include <linux/seq_file.h>
36 #include <linux/console.h>
37 #include <linux/mca.h>
38 #include <linux/root_dev.h>
39 #include <linux/highmem.h>
40 #include <linux/module.h>
41 #include <linux/efi.h>
42 #include <linux/init.h>
43 #include <linux/edd.h>
44 #include <linux/nodemask.h>
45 #include <linux/kexec.h>
46 #include <linux/crash_dump.h>
47 #include <linux/dmi.h>
48 #include <linux/pfn.h>
50 #include <video/edid.h>
54 #include <asm/mpspec.h>
55 #include <asm/mmzone.h>
56 #include <asm/setup.h>
57 #include <asm/arch_hooks.h>
58 #include <asm/sections.h>
59 #include <asm/io_apic.h>
63 #include <setup_arch.h>
64 #include <bios_ebda.h>
66 /* This value is set up by the early boot code to point to the value
67 immediately after the boot time page tables. It contains a *physical*
68 address, and must not be in the .bss segment! */
69 unsigned long init_pg_tables_end __initdata
= ~0UL;
71 int disable_pse __devinitdata
= 0;
76 extern struct resource code_resource
;
77 extern struct resource data_resource
;
79 /* cpu data as detected by the assembly code in head.S */
80 struct cpuinfo_x86 new_cpu_data __cpuinitdata
= { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
81 /* common cpu data for all cpus */
82 struct cpuinfo_x86 boot_cpu_data __read_mostly
= { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
83 EXPORT_SYMBOL(boot_cpu_data
);
85 unsigned long mmu_cr4_features
;
87 /* for MCA, but anyone else can use it if they want */
88 unsigned int machine_id
;
90 EXPORT_SYMBOL(machine_id
);
92 unsigned int machine_submodel_id
;
93 unsigned int BIOS_revision
;
94 unsigned int mca_pentium_flag
;
96 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
99 /* user-defined highmem size */
100 static unsigned int highmem_pages
= -1;
105 struct screen_info screen_info
;
106 EXPORT_SYMBOL(screen_info
);
107 struct apm_info apm_info
;
108 EXPORT_SYMBOL(apm_info
);
109 struct edid_info edid_info
;
110 EXPORT_SYMBOL_GPL(edid_info
);
111 struct ist_info ist_info
;
112 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
113 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
114 EXPORT_SYMBOL(ist_info
);
117 extern void early_cpu_init(void);
118 extern int root_mountflags
;
120 unsigned long saved_videomode
;
122 #define RAMDISK_IMAGE_START_MASK 0x07FF
123 #define RAMDISK_PROMPT_FLAG 0x8000
124 #define RAMDISK_LOAD_FLAG 0x4000
126 static char __initdata command_line
[COMMAND_LINE_SIZE
];
128 struct boot_params __initdata boot_params
;
130 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
132 #ifdef CONFIG_EDD_MODULE
136 * copy_edd() - Copy the BIOS EDD information
137 * from boot_params into a safe place.
140 static inline void copy_edd(void)
142 memcpy(edd
.mbr_signature
, EDD_MBR_SIGNATURE
, sizeof(edd
.mbr_signature
));
143 memcpy(edd
.edd_info
, EDD_BUF
, sizeof(edd
.edd_info
));
144 edd
.mbr_signature_nr
= EDD_MBR_SIG_NR
;
145 edd
.edd_info_nr
= EDD_NR
;
148 static inline void copy_edd(void)
153 int __initdata user_defined_memmap
= 0;
156 * "mem=nopentium" disables the 4MB page tables.
157 * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
158 * to <mem>, overriding the bios size.
159 * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
160 * <start> to <start>+<mem>, overriding the bios size.
162 * HPA tells me bootloaders need to parse mem=, so no new
163 * option should be mem= [also see Documentation/i386/boot.txt]
165 static int __init
parse_mem(char *arg
)
170 if (strcmp(arg
, "nopentium") == 0) {
171 clear_bit(X86_FEATURE_PSE
, boot_cpu_data
.x86_capability
);
174 /* If the user specifies memory size, we
175 * limit the BIOS-provided memory map to
176 * that size. exactmap can be used to specify
177 * the exact map. mem=number can be used to
178 * trim the existing memory map.
180 unsigned long long mem_size
;
182 mem_size
= memparse(arg
, &arg
);
183 limit_regions(mem_size
);
184 user_defined_memmap
= 1;
188 early_param("mem", parse_mem
);
190 #ifdef CONFIG_PROC_VMCORE
191 /* elfcorehdr= specifies the location of elf core header
192 * stored by the crashed kernel.
194 static int __init
parse_elfcorehdr(char *arg
)
199 elfcorehdr_addr
= memparse(arg
, &arg
);
202 early_param("elfcorehdr", parse_elfcorehdr
);
203 #endif /* CONFIG_PROC_VMCORE */
206 * highmem=size forces highmem to be exactly 'size' bytes.
207 * This works even on boxes that have no highmem otherwise.
208 * This also works to reduce highmem size on bigger boxes.
210 static int __init
parse_highmem(char *arg
)
215 highmem_pages
= memparse(arg
, &arg
) >> PAGE_SHIFT
;
218 early_param("highmem", parse_highmem
);
221 * vmalloc=size forces the vmalloc area to be exactly 'size'
222 * bytes. This can be used to increase (or decrease) the
223 * vmalloc area - the default is 128m.
225 static int __init
parse_vmalloc(char *arg
)
230 __VMALLOC_RESERVE
= memparse(arg
, &arg
);
233 early_param("vmalloc", parse_vmalloc
);
236 * reservetop=size reserves a hole at the top of the kernel address space which
237 * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
238 * so relocating the fixmap can be done before paging initialization.
240 static int __init
parse_reservetop(char *arg
)
242 unsigned long address
;
247 address
= memparse(arg
, &arg
);
248 reserve_top_address(address
);
251 early_param("reservetop", parse_reservetop
);
254 * Determine low and high memory ranges:
256 unsigned long __init
find_max_low_pfn(void)
258 unsigned long max_low_pfn
;
260 max_low_pfn
= max_pfn
;
261 if (max_low_pfn
> MAXMEM_PFN
) {
262 if (highmem_pages
== -1)
263 highmem_pages
= max_pfn
- MAXMEM_PFN
;
264 if (highmem_pages
+ MAXMEM_PFN
< max_pfn
)
265 max_pfn
= MAXMEM_PFN
+ highmem_pages
;
266 if (highmem_pages
+ MAXMEM_PFN
> max_pfn
) {
267 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn
- MAXMEM_PFN
), pages_to_mb(highmem_pages
));
270 max_low_pfn
= MAXMEM_PFN
;
271 #ifndef CONFIG_HIGHMEM
272 /* Maximum memory usable is what is directly addressable */
273 printk(KERN_WARNING
"Warning only %ldMB will be used.\n",
275 if (max_pfn
> MAX_NONPAE_PFN
)
276 printk(KERN_WARNING
"Use a HIGHMEM64G enabled kernel.\n");
278 printk(KERN_WARNING
"Use a HIGHMEM enabled kernel.\n");
279 max_pfn
= MAXMEM_PFN
;
280 #else /* !CONFIG_HIGHMEM */
281 #ifndef CONFIG_HIGHMEM64G
282 if (max_pfn
> MAX_NONPAE_PFN
) {
283 max_pfn
= MAX_NONPAE_PFN
;
284 printk(KERN_WARNING
"Warning only 4GB will be used.\n");
285 printk(KERN_WARNING
"Use a HIGHMEM64G enabled kernel.\n");
287 #endif /* !CONFIG_HIGHMEM64G */
288 #endif /* !CONFIG_HIGHMEM */
290 if (highmem_pages
== -1)
292 #ifdef CONFIG_HIGHMEM
293 if (highmem_pages
>= max_pfn
) {
294 printk(KERN_ERR
"highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages
), pages_to_mb(max_pfn
));
298 if (max_low_pfn
-highmem_pages
< 64*1024*1024/PAGE_SIZE
){
299 printk(KERN_ERR
"highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages
));
302 max_low_pfn
-= highmem_pages
;
306 printk(KERN_ERR
"ignoring highmem size on non-highmem kernel!\n");
313 * workaround for Dell systems that neglect to reserve EBDA
315 static void __init
reserve_ebda_region(void)
318 addr
= get_bios_ebda();
320 reserve_bootmem(addr
, PAGE_SIZE
);
323 #ifndef CONFIG_NEED_MULTIPLE_NODES
324 void __init
setup_bootmem_allocator(void);
325 static unsigned long __init
setup_memory(void)
328 * partially used pages are not usable - thus
329 * we are rounding upwards:
331 min_low_pfn
= PFN_UP(init_pg_tables_end
);
335 max_low_pfn
= find_max_low_pfn();
337 #ifdef CONFIG_HIGHMEM
338 highstart_pfn
= highend_pfn
= max_pfn
;
339 if (max_pfn
> max_low_pfn
) {
340 highstart_pfn
= max_low_pfn
;
342 printk(KERN_NOTICE
"%ldMB HIGHMEM available.\n",
343 pages_to_mb(highend_pfn
- highstart_pfn
));
344 num_physpages
= highend_pfn
;
345 high_memory
= (void *) __va(highstart_pfn
* PAGE_SIZE
- 1) + 1;
347 num_physpages
= max_low_pfn
;
348 high_memory
= (void *) __va(max_low_pfn
* PAGE_SIZE
- 1) + 1;
350 #ifdef CONFIG_FLATMEM
351 max_mapnr
= num_physpages
;
353 printk(KERN_NOTICE
"%ldMB LOWMEM available.\n",
354 pages_to_mb(max_low_pfn
));
356 setup_bootmem_allocator();
361 void __init
zone_sizes_init(void)
363 unsigned long max_zone_pfns
[MAX_NR_ZONES
];
364 memset(max_zone_pfns
, 0, sizeof(max_zone_pfns
));
365 max_zone_pfns
[ZONE_DMA
] =
366 virt_to_phys((char *)MAX_DMA_ADDRESS
) >> PAGE_SHIFT
;
367 max_zone_pfns
[ZONE_NORMAL
] = max_low_pfn
;
368 #ifdef CONFIG_HIGHMEM
369 max_zone_pfns
[ZONE_HIGHMEM
] = highend_pfn
;
370 add_active_range(0, 0, highend_pfn
);
372 add_active_range(0, 0, max_low_pfn
);
375 free_area_init_nodes(max_zone_pfns
);
378 extern unsigned long __init
setup_memory(void);
379 extern void zone_sizes_init(void);
380 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
382 void __init
setup_bootmem_allocator(void)
384 unsigned long bootmap_size
;
386 * Initialize the boot-time allocator (with low memory only):
388 bootmap_size
= init_bootmem(min_low_pfn
, max_low_pfn
);
390 register_bootmem_low_pages(max_low_pfn
);
393 * Reserve the bootmem bitmap itself as well. We do this in two
394 * steps (first step was init_bootmem()) because this catches
395 * the (very unlikely) case of us accidentally initializing the
396 * bootmem allocator with an invalid RAM area.
398 reserve_bootmem(__pa_symbol(_text
), (PFN_PHYS(min_low_pfn
) +
399 bootmap_size
+ PAGE_SIZE
-1) - __pa_symbol(_text
));
402 * reserve physical page 0 - it's a special BIOS page on many boxes,
403 * enabling clean reboots, SMP operation, laptop functions.
405 reserve_bootmem(0, PAGE_SIZE
);
407 /* reserve EBDA region, it's a 4K region */
408 reserve_ebda_region();
410 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
411 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
412 unless you have no PS/2 mouse plugged in. */
413 if (boot_cpu_data
.x86_vendor
== X86_VENDOR_AMD
&&
414 boot_cpu_data
.x86
== 6)
415 reserve_bootmem(0xa0000 - 4096, 4096);
419 * But first pinch a few for the stack/trampoline stuff
420 * FIXME: Don't need the extra page at 4K, but need to fix
421 * trampoline before removing it. (see the GDT stuff)
423 reserve_bootmem(PAGE_SIZE
, PAGE_SIZE
);
425 #ifdef CONFIG_ACPI_SLEEP
427 * Reserve low memory region for sleep support.
429 acpi_reserve_bootmem();
431 #ifdef CONFIG_X86_FIND_SMP_CONFIG
433 * Find and reserve possible boot-time SMP configuration:
438 #ifdef CONFIG_BLK_DEV_INITRD
439 if (LOADER_TYPE
&& INITRD_START
) {
440 if (INITRD_START
+ INITRD_SIZE
<= (max_low_pfn
<< PAGE_SHIFT
)) {
441 reserve_bootmem(INITRD_START
, INITRD_SIZE
);
442 initrd_start
= INITRD_START
+ PAGE_OFFSET
;
443 initrd_end
= initrd_start
+INITRD_SIZE
;
446 printk(KERN_ERR
"initrd extends beyond end of memory "
447 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
448 INITRD_START
+ INITRD_SIZE
,
449 max_low_pfn
<< PAGE_SHIFT
);
455 if (crashk_res
.start
!= crashk_res
.end
)
456 reserve_bootmem(crashk_res
.start
,
457 crashk_res
.end
- crashk_res
.start
+ 1);
462 * The node 0 pgdat is initialized before all of these because
463 * it's needed for bootmem. node>0 pgdats have their virtual
464 * space allocated before the pagetables are in place to access
465 * them, so they can't be cleared then.
467 * This should all compile down to nothing when NUMA is off.
469 static void __init
remapped_pgdat_init(void)
473 for_each_online_node(nid
) {
475 memset(NODE_DATA(nid
), 0, sizeof(struct pglist_data
));
480 static void set_mca_bus(int x
)
485 static void set_mca_bus(int x
) { }
488 /* Overridden in paravirt.c if CONFIG_PARAVIRT */
489 char * __init
__attribute__((weak
)) memory_setup(void)
491 return machine_specific_memory_setup();
495 * Determine if we were loaded by an EFI loader. If so, then we have also been
496 * passed the efi memmap, systab, etc., so we should use these data structures
497 * for initialization. Note, the efi init code path is determined by the
498 * global efi_enabled. This allows the same kernel image to be used on existing
499 * systems (with a traditional BIOS) as well as on EFI systems.
501 void __init
setup_arch(char **cmdline_p
)
503 unsigned long max_low_pfn
;
505 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
506 pre_setup_arch_hook();
510 * FIXME: This isn't an official loader_type right
511 * now but does currently work with elilo.
512 * If we were configured as an EFI kernel, check to make
513 * sure that we were loaded correctly from elilo and that
514 * the system table is valid. If not, then initialize normally.
517 if ((LOADER_TYPE
== 0x50) && EFI_SYSTAB
)
521 ROOT_DEV
= old_decode_dev(ORIG_ROOT_DEV
);
522 screen_info
= SCREEN_INFO
;
523 edid_info
= EDID_INFO
;
524 apm_info
.bios
= APM_BIOS_INFO
;
526 saved_videomode
= VIDEO_MODE
;
527 if( SYS_DESC_TABLE
.length
!= 0 ) {
528 set_mca_bus(SYS_DESC_TABLE
.table
[3] & 0x2);
529 machine_id
= SYS_DESC_TABLE
.table
[0];
530 machine_submodel_id
= SYS_DESC_TABLE
.table
[1];
531 BIOS_revision
= SYS_DESC_TABLE
.table
[2];
533 bootloader_type
= LOADER_TYPE
;
535 #ifdef CONFIG_BLK_DEV_RAM
536 rd_image_start
= RAMDISK_FLAGS
& RAMDISK_IMAGE_START_MASK
;
537 rd_prompt
= ((RAMDISK_FLAGS
& RAMDISK_PROMPT_FLAG
) != 0);
538 rd_doload
= ((RAMDISK_FLAGS
& RAMDISK_LOAD_FLAG
) != 0);
544 printk(KERN_INFO
"BIOS-provided physical RAM map:\n");
545 print_memory_map(memory_setup());
550 if (!MOUNT_ROOT_RDONLY
)
551 root_mountflags
&= ~MS_RDONLY
;
552 init_mm
.start_code
= (unsigned long) _text
;
553 init_mm
.end_code
= (unsigned long) _etext
;
554 init_mm
.end_data
= (unsigned long) _edata
;
555 init_mm
.brk
= init_pg_tables_end
+ PAGE_OFFSET
;
557 code_resource
.start
= virt_to_phys(_text
);
558 code_resource
.end
= virt_to_phys(_etext
)-1;
559 data_resource
.start
= virt_to_phys(_etext
);
560 data_resource
.end
= virt_to_phys(_edata
)-1;
564 if (user_defined_memmap
) {
565 printk(KERN_INFO
"user-defined physical RAM map:\n");
566 print_memory_map("user");
569 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
570 *cmdline_p
= command_line
;
572 max_low_pfn
= setup_memory();
576 * Must be after max_low_pfn is determined, and before kernel
577 * pagetables are setup.
583 * NOTE: before this point _nobody_ is allowed to allocate
584 * any memory using the bootmem allocator. Although the
585 * alloctor is now initialised only the first 8Mb of the kernel
586 * virtual address space has been mapped. All allocations before
587 * paging_init() has completed must use the alloc_bootmem_low_pages()
588 * variant (which allocates DMA'able memory) and care must be taken
589 * not to exceed the 8Mb limit.
593 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
596 remapped_pgdat_init();
601 * NOTE: at this point the bootmem allocator is fully available.
604 paravirt_post_allocator_init();
608 #ifdef CONFIG_X86_GENERICARCH
609 generic_apic_probe();
616 * Parse the ACPI tables for possible boot-time SMP configuration.
618 acpi_boot_table_init();
622 #ifdef CONFIG_X86_IO_APIC
623 check_acpi_pci(); /* Checks more than just ACPI actually */
630 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
632 printk(KERN_WARNING
"More than 8 CPUs detected and "
633 "CONFIG_X86_PC cannot handle it.\nUse "
634 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
637 #ifdef CONFIG_X86_LOCAL_APIC
638 if (smp_found_config
)
642 e820_register_memory();
643 e820_mark_nosave_regions();
646 #if defined(CONFIG_VGA_CONSOLE)
647 if (!efi_enabled
|| (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
648 conswitchp
= &vga_con
;
649 #elif defined(CONFIG_DUMMY_CONSOLE)
650 conswitchp
= &dummy_con
;